pppd-2.3.10 bug?

[Alan Yates]

Hi,

I've been using pppd for a long time in a legacy analog dialup pool for
my ISP. I recently changed the way accounting information was
collected, using the /etc/ppp/auth-* scripts to log to a database rather
than the ipfwadm and cron script hack we used in the past.

Often the BYTES_SENT and BYTES_RCVD variables are 0, even after
considerable modem online time and traffic transfer. It works perfectly
most of the time, and CONNECT_TIME is always correct.

It appears that if a ppp connection is dropped without the remote end
closing 'nicely', ie sending a LCP termination request, by the time the
auth-down script is run the BYTE_* counters are cleared. NT peers in
particular appear to just drop the link not do normal LCP stuff.

I have emailed the pppd developers (Paul Mackerras et al in July this
year) with no response. Does anyone here know about this issue? I have
looked at the code, but don't have the testing enviroment (or a good
understanding of how it actually works) to trace the accounting code.

Alan

--
Alan Yates <al...@ay.com.au> http://www.alanyates.com/

[SlackGuy]

Thanks for the warning :) I currently use small perl scripts as
/etc/ppp/ip-up and down to set and read ip accounting rules using
ipchains, and then send the data to a RADIUS server. I was thinking
of using the BYTES_SENT in auth-up but now I'll leave well alone!